A Deep Second-Order Stochastic Residual Method for Fully Nonlinear Parabolic PDEs

DOI10.48550/arXiv.2607.16730MaRDI QIDQ7284179arXivFDO

Authors Zhen-Hua Zhao, Jihao Long, Zhenhua Zhao

Publication date 18 July 2026

Published in arXiv

Tldr We introduce D2SRM for high-dimensional fully nonlinear parabolic PDEs. Our main contribution is a novel second-order training scheme that jointly learns derivative-consistent approximations of the solution, gradient, and Hessian using Brownian one-step residuals and terminal value and gradient penalties.

Full work available at URL https://doi.org/10.48550/arXiv.2607.16730

arXiv classification

math.NAmath.OCstat.ML

This page was built for publication: A Deep Second-Order Stochastic Residual Method for Fully Nonlinear Parabolic PDEs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284179)